Mistakes

Below serves like a training history in ML, to minimise the cost.

Courses

If time consuming in building the model, use subset

If subset for sanity check on whether the program works or not first before anything else.

Understand the questions

“how”, “discuss” and “explain reasons” are different.

Self

Single Responsibility Principle

Should use multiple files for multiple purpose. E.g. I shouldn’t have put the markdown syntax cheatsheet to contain Jekyll specific syntax thou they are related.

Organise TODO

Should have one system e.g. google task to control all the TODO task instead of using ##TODO in the coding files (or bits here bits there). Otherwise, forget.

Assumptions

Should make as little assumption as possible.
E.g. the permenant link design for this blog is based on filename with days, but if I make file with same name on the same day, errors occur.This is also why Navie Bayes is fast but less accurate.

Logs

Carry on a logs like google calendar allows faster backtrack and reflection/better organisation.

Categorise

E.g. Whether I want to put each UNSW course under heading of different subject e.g. Machine Learning or alternatively, the one currently used, is to categorise them under one title “UNSW courses”, and put others in e.g. time series, numerical, text, classification.

Project

Teamwork

When two or more work on the same problem, communicate, daily, hourly to avoid duplicate of work.

Specialisation, let the ppl who fit the task well do it, save time & struggles.

Framework

It’s worth to research and find the optimal framework for the project. E.g. In one project course, we initially choose VM to run ROS, which is slow but compatiable with all OS. 2 Terms later, we realise ROS can be run one WSL2 for window, which is faster and simplier. But by that time, it is already quite late since my teammates busy with other school work and decide to stick with VM.

Community

When go into a new subject / place, join community earlier, more timely information always better. Even better find ppl who start at the same time - a frame of reference.